The transformation of the semiconductor industry in the era of data explosion

As we all know, in the past five to ten years, the science and technology industry has undergone tremendous changes. The emerging new technology not only makes our life easier and more convenient, but the frequent use of smart devices also makes our IQ more and more. "This improvement in IQ is different from the improvement brought by previous education. It is not the improvement of aristocraticization, but the increase of civilians and the general IQ."

On the other hand, the emergence of social software such as WeChat makes the distance between any two people in the world shorter than ever. It has been said that any two people in the world can be connected through social networks. This distance was 5.3 people five years ago, and with the popularity of social networks, the distance has been shortened to 3.5 people.

The era of data explosion, the transformation of the semiconductor industry

For ordinary people, this means that the actual communication between people is more tight, the distance is shortened, and communication is more convenient. What does this technology change mean for equipment manufacturers? How does the semiconductor industry complete the transformation in such technological changes? At MWC2017, Green Core CEO Sanjay Jha made his point.

Big data explosion, new architecture becomes inevitable

In Sanjay Jha's view, the main effect of this change is twofold.

First of all, there is an increase in the variety of testing equipment. For example, cameras have many different social forms. Under the influence of the Internet of Things wave, in order to achieve more functions and data collection, these detection devices must be connected to the network. Now we can see a lot of monitoring equipment in the market that can detect parameters such as temperature and light brightness. All need to be connected to the network. Without the data support of the back end, the function of the front end does not have much significance.

Another notable change is the increasing consumption and amount of data on wireless networks. “When we talked about wireless network data with our customers a few years ago, our concept of data mainly stayed in applications such as pictures and audio. But with the advent of the iPhone and the advent of smartphones, we’ve found that data usage is It has grown geometrically. Whether it is the application of 3G network a few years ago or the 4G used now, or the application of the upcoming 5G network. The data transmission speed is getting faster and faster, and the requirements for equipment are getting higher and higher. "Sanjay Jha said.

The first impact of this exploding data with geometric progression is the data center, which is getting faster and faster, and the data stored and exchanged in the data center is getting faster and faster.

In the original mobile phone application, the connection between people was often one-to-one, that is to say mainly through telephone text messages. However, the value of the smartphone and the application of the network of the three systems are not urgent. The application scores of data are becoming more diverse.

We have noticed that the application of data is no longer limited to pictures and audio, but also to a variety of applications such as video.

Taking smart cars as an example, if the car is sufficiently intelligent in the near future, based on real-time analysis of the car conditions and driving behavior, the system can actively intervene in such dangerous driving, which can completely eliminate such accidents. And all of this is based on the car data behind the car's bus technology. Automobile data is the basic requirement of the entire automotive aftermarket ecosystem. Complete vehicle data collection, analysis and analysis technologies are the foundation of smart cars in the future.

However, the amount of data generated by smart cars during driving is completely different from traditional applications. How to solve the problem of data processing? In Sanjay Jha's view, artificial intelligence is needed at this time for data processing.

“In fact, artificial intelligence technology has different forms in different applications. Artificial intelligence can be a chip specially used for graphics processing. Such chips do a good job of parallel computing and are suitable for neural analysis. Artificial intelligence can also be professional. Chips, these chips are more suitable for deep learning. Which form is it depends on the actual application."

However, we should also note that whether it is neuroanalysis or deep learning, it requires an different architecture different from traditional applications, which has become an inevitable development trend.

Moore's Law has been broken? Just the law of economics

At the end of all things, especially in the unbalanced competition of the PC chip industry in recent years, the sound of Moore's Law is inexhaustible. This is another change brought about by the development of science and technology.

"In essence, Moore's Law is a phenomenon of economics. Physics, especially the semiconductor industry, is the foundation of Moore's Law and leads the competition in economics," said Sanjay Jha.

The existence of Moore's Law has greatly accelerated the development of science and technology from the 20th to the 21st century, and is even more manifested in our lives. E-commerce, virtual reality, online communities, office automation, smart homes... Of course, everything around you is taken for granted. In the past, it was impossible to understand.

However, in the context of the slowdown in Moore's Law and the decline in the speed of semiconductor technology, it is very challenging to provide more computing power.

Looking at the history of the semiconductor industry: every decade in the past has been driven by new semiconductor technologies. In the 1970s, it was MOS tube memory, which made TI, Fairchild, and IR; in the 1980s, it was a microprocessor, which made NEC and Motorola; in the 1990s, it was CPU and memory, Intel and Toshiba skyrocketed; after 2000, wireless interconnection and Fabless rose, Qualcomm Start to rise. Advances in technology have led to product updates, which have driven the industry's transformation in nature.

However, to this day, Moore's Law is gradually slowing down, and the technological development of semiconductors seems to have reached a crossroads. We have reached the level where the miniaturization of transistors is close to physical limits.

The new semiconductor technology node will continue to bring significant miniaturization and low power consumption in the next decade, but the cost is increasing and the historical improvement in speed has not been realized.

How to solve this problem and adapt to the changes brought about by technology? SanjayJha believes that this requires us to be able to change our thinking and find another solution. After all, "potential" is artificial!

FD-SOI, another way of thinking about "potential"

So where is the other idea?

In February 2017, Grofund and Chengdu Municipality of Sichuan Province, China, jointly formed a core 12-inch plant, with a total investment of 9.053 billion US dollars.

In May, Gexin once again announced that it will cooperate with Chengdu to jointly implement the FD-SOI Ecosphere Action Plan and establish a world-class FD-SOI ecosystem in Chengdu.

It is reported that the new plant of Green Core Chengdu will build the world's first 22nm FD-SOI process production line, and the products will be widely used in mobile terminals, Internet of Things, automotive electronics and other fields.

The first phase will build a 12-inch wafer production line for mainstream CMOS processes. The main process is 0.18um/0.13um. The related process technology will be transferred from Singapore. Some of the equipment will come from Singapore. It is expected to be put into operation at the end of 2018. Ten thousand pieces. In the second phase, the latest 22FDX22nmFD-SOI process 12-inch wafer production line will be built. The related process will be transferred from Fab1 in Dresden, Germany in the second half of 2018. The equipment will be newly purchased and is expected to be put into production in the fourth quarter of 2019. The monthly production capacity is 65,000 pieces. Gexin will establish an ecosystem with Chengdu and is expected to introduce a more complete IP\EDA design service for Chengdu to better support the 22FDX system.

In addition to the mainstream FinFET technology, FD-SOI (full-discharge silicon-on-insulator) technology has also received increasing attention in the wave of the Internet of Things. Compared with FinFET, FD-SOI substrate is more expensive, but the number of masks and manufacturing process is less than that of FinFET. The reduction of part of the mask cost also reduces the manufacturing time. It is technically more suitable for analog/mixed signal and RF than FinFET. FD-SOI also has the advantage of low power consumption. Therefore, SanjayJha believes that FD-SOI will be a better choice for IoT, or it can complement FinFET.

In Sanjay Jha's view, even in the smartphone market, FD-SOI has a good job.

A good example is that MediaTek recently began to evaluate the adoption of the FD-SOI process, which is extremely hot in the mainland market, after the failure of the 10nm FinFET process technology. MediaTek opened its market in the middle and low-end mobile phone chips. Spreadtrum is struggling, considering the latest 22-nanometer FD-SOI process.

Supply chain industry pointed out that MediaTek's internal evaluation in the core of a 22-nm FD-SOI process chip, the main low-end mobile phone chip market, which will be 28 nanometer planar transistor (PlanarCMOS) and 16/14 nanometer FinFET two Another new battlefield beyond big technology, and avoiding the disadvantages of high power consumption of Planar CMOS and high cost of FinFET, using the low power consumption, low leakage and low cost advantages of FD-SOI process, creating a new field in the middle and low-end mobile phone chip field .

Localization and deepening the potential of the Chinese market

As one of the three largest wafer foundries in the industry, Groffont has been promoting FD-SOI technology for several years.

According to IC Insights' forecast data, in 2016, Gexin still firmly occupied the second place in the world with an 11% market share, with revenue of 5.545 billion US dollars. In comparison, the global foundry leader TSMC's revenue was 29.488 billion US dollars, with a market share of 59%; the third place Taiwan's Lianhua Electronics revenue was 4.582 billion US dollars, with a market share of 9%.

The reason why the 12-inch fab in China did not choose the 40nm/55nm process is because the core has seen some 40nm demand shift to 22nm FDX. The grid factory in Singapore has a production of 40nm/55nm, so according to demand forecast, the core selection of 0.18um/0.13um on the mature process.

In addition, how to localize in China and tap the potential of the Chinese market is also a problem that Gexin has to think about. In Sanjay Jha's view, building factories in China and even building an ecosystem with Chengdu is an indispensable step for localization.

“How do we root the global trend in the soil of China? This is a question we have been thinking about,” SanjayJha said. “In recent years, we have noticed that the Chinese market has a lot of emerging innovations and Chinese manufacturing. In China, we have seen China’s innovation, witnessed China’s massive investment in artificial intelligence, and participated in many investments. But we hope more In-depth participation in it."

Establishing a good partnership with the Chengdu government to create an FD-SOI ecosystem system and network is one aspect. On the other hand, Gexin also hopes to attract more investment and attract all people with lofty ideals to participate in the ecosystem. In the construction of the joint efforts to create innovative products.

to sum up

For Gexin, the word “potential” is artificial, which is probably the best depiction of the driving force for the future development of Gexin.

In the view of Gexin, grasping the impact of technological changes on the semiconductor industry and promoting the development of intelligence under the premise of interconnection, we need to pay attention to the following elements: First, low power consumption. For non-standard networks, especially for neural networks, low power consumption is very important. Followed by memory. Through the cooperation between memory and bandwidth, we can achieve better memory and processing speed, which can also reduce power consumption.

Whether it is artificial intelligence or other technologies that will emerge in the future, there is a basic condition that we must use the most basic and specific technology to support its operation. The fab is the core of all infrastructure.

“So we are here to hope that our business in China will be successful and our business in the world will be successful. As a leading company in fab, wafer fabrication and foundry, we hope to cooperate with you. Let artificial intelligence become a reality, and be able to enter the real next generation of "potential" in the artificial artificial intelligence world!" SanjayJha said.

Editor in charge: Malan

Sherpa Lined Slipper Socks

Cozy Double Lined slipper Booties Christmas Soft warm ballerina slippers are made with a soft pom pom sherpa lining with a colorful bling sequin exterior. Snoozies are soft, comfy, warm and just simply adorable – the newest foot coverings for your feet! Each pair has a non skid sole so you can feel safe wearing them on wood or tile floors. These Slipper Socks keep feet warm even on the coldest of days! They work in warm climates too where air conditioning keeps floors cold. They are are fully machine washable and won`t shrink!

Sherpa Lined Slipper Socks,Sherpa Sock Slippers,Sherpa Fuzzy Socks,Mens Sherpa Lined Slipper Socks

jiangyin haorun knitting co.,ltd , https://www.jyhrknitting.com